Machine learning assisted materials design of high-speed railway wheel with better fatigue performance

Abstract

To improve the fatigue fracture resistance of high-speed railway wheels, this paper studies the matching optimization design of microscopic crystal parameters of wheel materials based on machine learning methods. The microscopic crystal parameters of a pearlitic steel of high-speed railway wheel are collected, the relationship between microscopic crystal parameters and material properties of wheel material is studied. Based on machine learning methods including artificial neural network, random forest, support vector machine and K-nearest neighbor, the influence of the microscopic crystal parameters on the material properties of wheel steel is studied. The results show that the artificial neural network has a higher prediction accuracy, the microstructure parameters have a complex effect on fatigue performance of the wheel. When the prior austenite grain size of the structure is 55–65 μm, the pearlite colony size is 27–29 μm, and the pearlite interlamellar spacing is 95–105 nm, the fatigue resistance of the wheel steel is the best, and the fatigue strength is 528 MPa. The optimization result meets the requirements of high-speed railway wheel strength evaluation standard with a higher safetyfactor by finite element simulation.